A Hybrid 3D-CNN and BiLSTM Framework for Early Alzheimer's Disease Prediction Using Neuroimaging and Clinical Biomarkers

Alzheimer's disease (AD) is the most common cause of dementia, accounting for 60-70% of cases and affecting more than 55 million people worldwide. Early and accurate differentiation of AD, Mild Cognitive Impairment (MCI) and Cognitively Normal (CN) subjects is essential for timely intervention. Here, we propose a hybrid deep learning framework that integrates a 3D Convolutional Neural Network (3D-CNN) and a Bidirectional Long Short-Term Memory (BiLSTM) network to jointly learn spatial atrophy patterns from structural MRI and temporal trends from longitudinal cognitive and cerebrospinal fluid (CSF) biomarkers. Feature-level late fusion combines both representations before classification. The framework achieved an accuracy of 94.8%, sensitivity of 93.6%, specificity of 95.9%, F1-score of 94.3% and AUC of 0.971 with 600 subjects from ADNI database, which outperformed the SVM, Random Forest and standalone 3D-CNN/BiLSTM baselines. The ablation studies and Wilcoxon signed-rank tests (p < 0.01) validate the statistically significant gain by combining spatial and temporal information, supporting the potential of the framework for automated three-class AD screening.